The Internet began as a simple database of limited textual information, and quickly transformed into an extensive database of images, text, and audio information. It would take several lifetimes to hunt for various kinds of information throughout the Internet and USENET news groups, and, all the while, the number of files would be expanding faster than anyone's ability to peruse them.
Search engines were devised to manage the hunt. Search engines are programs that search the Internet for documents that contain specified keywords and return a list of documents which contain those keywords. These engines run programs called “spiders” that continuously explore the Internet and, often, USENET news groups, they index the information on websites that the spiders encounter. Indexing forms a vast database of website addresses that are associated with key words that have been found on the websites themselves.
Search engines such as Yahoo, Google, MSN, and International Business Machines' CLEVER require the user to enter at least one key term or query into a text field. Keywords, phrases, phrases in quotes, and Boolean queries are matched to various sites on the Internet, and when the query is complete a list of these sites is displayed for the user's review.
Although the most widely used search engines have a category that enables them to access images, none of them allows an image to be entered as a query or search entity. All known engines require that the user enter a text query, and the search hits files that display images that are associated with the entered text query. If a person sees an image and wishes to access online information about it, he or she will have to search for it using a text query. The user cannot use the image itself as a query. If the user cannot put his or her search request into words, he or she will not be able to conduct a search in a standard online search engine.
Several innovators are working to solve this need. Hewlett-Packard, for example, has developed a method of indexing an image that is based on information derived from a global positioning system (GPS). The system obtains an image along with its location, and indexes images according to their location. Such systems are useful in organizing album data since some digital cameras can acquire GPS data and correlate it with captured imagery. However, searching is limited to images that have a significant correlation with a given location.
A search engine developed by Xerox Corporation incorporates a multi-modal browsing and clustering system to retrieve image data. The system seeks similarities between images not only in textual references, but also in other associated information such as in-links, out-links, image characteristics, text genre, and the like. However, this engine is limited to specific image types which have defined colors, contain text, and have other visual identifiers. In short, the Xerox engine requires the images to have such specific characteristics, it limits the system's utility and viability as an all-purpose search engine.
Some attempts have been made to extract information from databases using images themselves as search entities rather than keywords related to the images. These systems can translate, provide information about, or interpret objects contained in an image. These systems generally work as follows. An input device extracts the object of interest from its background. The object is compared with objects stored in a pre-populated database to find a match. Finally, the system retrieves information in the database about the object and permits it to be displayed to the user. However, the system is limited to images containing extractable, defined objects, such as fruits, articles, animals, or any object which is easily outlined. However, many images require identification as a whole entity, such as an image of a geographic locations or a piece of artwork. As a result, this method has limited applicability.
Complex images with a myriad of superfluous objects are easier to identify using methods such as pixel analysis. Using this method, a database is populated with primitive, weighted vectors of images that facilitate the image processing. The inputted images are compared and matched through specific vectors that define them. Therefore, there remains a clear need for a system capable of capturing images, converting those images into computer readable formats, using the processed images as search queries in a search engine, comparing the images to images stored in the database, and, upon finding a match, displaying information associated with the image to a user of the system.